/Louvain Brain · hosted Coming soon

A brain for your company, running in seconds.

Louvain Engine, run for you: one brain per company, configured lane by lane, with permissions on a dial. There’s no app on top. Your agents and your software ask it.

Brains are opening to early customers first. Talk to us for early access.

01Lanes

Configure the brain the way it works: lane by lane. Each lane is a setting you can see and change, and every change says what it differs from.

Your model: tag contracts

Your own model and instructions, as a step: here, tag each contract with its party, value and renewal date.

Run it on one document before it runs on the rest.

  1. ingest

    Ingestion

    Your pipeline as a process map you edit. Plug in your own model and your own instructions as a step, and test any step on a real document before it runs on the rest.

  2. embed

    Embedding

    The embedder the benchmark was scored with, or your own. One brain, one space, chosen when it starts.

  3. rerank · keyword

    Recall

    The reranker that reads whole documents, and keyword search with an exact field for names and codes.

  4. models

    Models

    Which model reads, at which step, on whose key. Bring your own provider, or open weights end to end.

02Permissions on a dial

As open as one team, or as strict as your sources. Start open, and turn it up as more people and more documents arrive. Nobody reads past their rung.

Level 3 of 4

Your app says who

Your software says who is asking, and the brain returns only what that person may read.

A sales rep’s agent asks. It can read:

  • Sales playbook: can read
  • Board pack, Q3: can read
  • Salary bands: cannot read

03The same engine

Start self-hosted, move by changing a URL.

A brain speaks the open engine’s tools, routes and settings. The same MCP tools, the same ingest format and the same setting names, so an agent or a feed built on Louvain Engine points at a brain with a new URL and a new key.

What a brain adds is what a company needs once more than one team reads it: permissions, sources kept in step, and a control plane that runs one brain per company for you.

Pricing

A brain is priced for your company: by its documents, its people and how it’s configured. Talk to us and we’ll quote it.

Talk to us

04One engine, three ways

Run it yourself, have us run it, or use Louvain.

  • Open source

    Louvain Engine

    You run it

    The open retrieval engine behind Louvain: it finds the right document in your company’s files, behind a permission fence, and hands it to your agent with its source, self-hosted, over MCP and HTTP.

    • A permission fence first in every query
    • Keyword and meaning search, with an optional reranker
    • Three MCP tools, an HTTP ingest door and a Search and Ask app
    • Plugins as functions in one file
    See the engine
  • Hosted · coming soon

    Louvain Brain

    We run it, one per company

    Louvain Engine, run for your company: a brain in seconds, configured lane by lane, with permissions on a dial.

    • Everything in the engine
    • Lanes: ingestion, embedding, recall, models
    • Permissions on a dial
    • No app on top: your agents and software ask it
    You are here
  • The product

    Louvain

    Ready to use

    The home screen of work: what was said, what was agreed and what is owed, with an answer behind every claim.

    • Built on the same engine
    • Connected to your email, files, chat and calendar
    • Today, the Plate and Ask, with receipts
    • Each person sees only what they may read
    See Louvain

Tell us what your brain should know.

We use Google Analytics, and Google’s and LinkedIn’s advertising tags, to see how this site is used and which ads bring people here. They set cookies only if you accept. Cookie Notice.